The Impact of Exchange Rate Volatility on the European and U.S. Labor Markets

The Impact of Exchange Rate Volatility on the European and U.S. Labor Markets Ansgar Belke University of Hohenheim Leo Kaas1 University of Vienna Abstract. We analyze the effects of exchange rate volatility on labor market performance, both theoretically and empirically. We consider a simple Dixit/Pindyck-style model to show that there is a negative relationship between exchange rate variability and job creation. It turns out that a higher reservation wage, a better bargaining position of workers and higher costs of job creation strengthen the adverse impact of uncertainty on employment. Thus, the link between exchange rate variability and employment should be stronger in most European countries than in the US. Our regression analysis provides support for this conclusion. JEL classification: E24, F33, F41 Keywords: Exchange rate variability; Job creation; Option value effects Manuscript received: December 21, 2003; Accepted: July 16, 2004 1. Introduction In past years the euro-dollar exchange rate, much as the DM-dollar rate before, has undergone large and unpredictable movements which are at times difficult to understand and which are often perceived to be politically costly. Mundell (2000, 2000a), for instance, argues that movements of the euro-dollar rate comparable to those of the DM-dollar rate since 1971 would break Euroland apart. There are different reasons why politicians and economists care about exchange rate variability. In the first place, it is typically argued that exchange rate variability discourages trade. However, a large empirical literature on this issue does not confirm a significant effect of exchange rate variability on the volume of trade.2 On the other hand, there is no compelling reason why the volume of trade should be a politically important variable in itself. In fact, what policy makers should care about is economic welfare, and indeed there is a large literature on the impact of exchange rate volatility on welfare (see, for instance, Chinn and Miller, 1998, Devereux and Engel, 1998, and Neumeyer, 1998). In this paper we focus on a subset of this broader issue by investigating the implications of exchange rate volatility for the labor market.

21

22

THE JOURNAL OF ECONOMIC ASYMMETRIES

DECEMBER 2004

How can exchange rate variability have a significant impact on labor markets, given the weak empirical link between exchange rate variability and the volume of trade mentioned above? Our answer to that question is that an increase in exchange rate variability may well have an immediate impact on job creation decisions of firms and therefore be reflected in the employment data, whereas there need not be a short-run impact on the volume of trade. The decision of a firm whether or not to invest (and to create jobs) in export-oriented activities incurs sunk costs, such as creating a new production line or building up a distribution system in foreign markets. Therefore, an increase in exchange rate volatility may well deter firms from creating employment, but firms who are already active in foreign trade will not cut their exports just because of an increase in exchange rate volatility. Another reason why exchange rate variability might not have an immediate impact on the volume of trade comes from the 'pricing to market' idea, that is, firms keep local prices fixed even in the face of large exchange rate changes.3 This implies that foreign sales should react little to exchange rates. Firms keep producing but their domestic currency earnings become variable whereas their domestic costs remain stable. Exchange rate variability can certainly influence the variability of profits, even if trade changes only a little. Therefore, firms might react to an increase in exchange rate (and hence profit) variability in the first instance by reducing investment and employment in trade-related activities. This might depress future trade volumes but firms will not necessarily export less in the short run just because exchange rate variability has increased. The long-run response, on the other hand, is more difficult to isolate in empirical work because there are other long-run trends that influence trade volumes (for example, reduction in transport costs, shifts in tastes, etc.) and because variability changes so much over time. The goal of this paper is twofold. First, we develop a simple model to illustrate a mechanism that explains a negative relationship between exchange rate uncertainty and job creation. The model is based on the idea that uncertainty of future earnings raises the ‘option value of waiting’ (see Dixit, 1989). When firms create a job, they have to incur sunk costs, such as hiring costs as well as the provision of job-specific capital. Moreover, wage payments are typically also sunk since firing restrictions and employment contracts prevent firms from laying off workers too rapidly. If the exchange rate is uncertain, firms fear an unfavorable appreciation of the (domestic) currency in which case they incur heavy losses. With high uncertainty and with binding employment contracts, firms may prefer to delay job creation, and this is so even if they are risk-neutral. On the other hand, even when non-binding contracts can be signed and work relationships can be terminated easily, higher volatility still has an adverse impact on employment via job destruction. Moreover, the better the bargaining position of workers is, the higher the option value of waiting and the stronger is the impact of uncertainty on employment. Since generous unemployment compensation systems, union power and firing restrictions generally improve the bargaining position of workers, the model predicts

VOL. 1 NO. 2

BELKE AND KAAS: EXCHANGE RATE VOLATILITY

23

that the link between exchange rate uncertainty and employment should be stronger in most European countries than in the US. The second goal of the paper is to provide some casual empirical evidence on the negative relation between exchange rate uncertainty and labor markets. We consider the influence of two measures of external exchange rate variability of the euro area4 and of the US on two key labor market indicators: (changes in) unemployment rates and employment growth.5 We find that exchange rate variability has a statistically significant negative impact on unemployment and employment in the euro area, even when adding various control variables. For the US, the evidence points only to an impact on unemployment, and the coefficients are typically smaller (though significant) than in the euro area. These results confirm the theoretical presumption that there is a negative impact of exchange rate variability on (un)employment which is more pronounced in the euro area where labor markets are perceived to be more rigid than in the US. The literature provides other mechanisms through which uncertainty may have an adverse impact on employment. First, in unionized labor markets in which contract wages are set in advance, uncertainty in labor demand (coming from uncertainty in productivity or in the exchange rate) may cause rational unions to set a higher wage than would otherwise be the case. Uncertainty results in a ´risk premium` in the wage, and thus in higher unemployment (Andersen and Sorensen, 1988, and Sorensen, 1992). In our paper such a mechanism is absent because employers assume all the risk. Another channel by which uncertainty might affect employment is via its impact on investment. Our theoretical arguments are equally valid for firms who decide on an investment project and, by the same reasoning, high uncertainty might induce firms to postpone investment projects (see Belke and Gros, 2001).6 Unemployment can be expected to rise if investment falls because investment is an important component of demand. Moreover, technological complementarities between labor and capital imply that a capital slowdown entails a fall in employment (see Rowthorn, 1999). The outline of the paper is as follows. In section 2, we develop our model of job creation and uncertainty to illustrate the negative relationship between uncertainty and employment. Section 3 defines our measure of exchange rate variability. Section 4 presents and comments on the regression results. Section 5 concludes. 2. Exchange rate uncertainty and employment In this section, we present a simple model of job creation and exchange rate uncertainty to illustrate the basic idea underlying the 'option value of waiting' following Dixit (1989) and Pindyck (1991). The model does not pretend to be close to reality. It is designed to convey the basic idea in a simple way. Moreover, our intention is to present a model that allows us to ask whether even a temporary, shortrun increase in uncertainty can have a strong impact on employment, and how this impact depends on features of the labor market.

24

THE JOURNAL OF ECONOMIC ASYMMETRIES

DECEMBER 2004

The model has three periods (called zero, one and two) and in period zero a single firm active in an export-oriented industry decides about job creation. During the first two periods the firm can create a job, hire a worker and produce output that is sold in a foreign market during the following periods. If the job is created during period zero, the worker is hired for two periods (zero and one) to produce output to be sold in periods one and two. If the job is created in period one, the worker is hired only for period one and output is sold in period two. To create a job, the firm pays a start-up cost c which reflects the cost of hiring, training and the provision of job-specific capital. After a job is created, a worker is hired and is paid a wage w above the worker’s fallback (or reservation) wage w during every period the worker is employed. The fallback wage measures (besides disutility of work) all opportunity income that the worker has to give up by accepting the job. In particular, it includes unemployment benefits, but it might also be positively related to a collective wage set by a trade union or to a minimum wage, both of which should raise the worker’s fallback position. In general we would argue that the fallback wage should be higher in countries that are characterized by generous unemployment benefit systems, by strong trade unions or by minimum wage legislation. In every period in which the worker is employed, he produces output to be sold in the following period in a foreign market at domestic price p which has a certain component p* (the foreign price) plus a stochastic component e (the exchange rate). We assume that the foreign price is fixed (‘pricing to market’), and that the exchange rate follows a random walk.7 In period one, the exchange rate e1 is uniformly distributed between –σ1 and +σ1. The exchange rate in period two, e2, is uniformly distributed between e1–σ2 and e1+σ2. An increase in σi means an increase in uncertainty, or an increase in the mean preserving spread in period i=1,2 (σi is proportional to the standard deviation of ei). Uncertainty can be temporary (if σ1>0 and σ2=0) or persistent (if also σ2>0). As will soon become apparent, however, the variability of the exchange rate during the second period has no influence on the outcome.8 The wage rate w for the job is determined by the (generalized) Nash bargaining solution that maximizes a weighted product of the worker’s and the firm’s expected net return from the job. We assume that both the firm and the worker are risk-neutral. This assumption implies that risk-sharing issues are of no importance for our analysis. Thus we may assume realistically (but without loss of generality) that the worker and the firm bargain about a fixed wage rate w (which is independent of realizations of the exchange rate) when the worker is hired, so that the firm bears all the exchange rate risk. A wage contract which shifts some exchange rate risk to the worker would leave the (unconditional) expected net returns unaffected, and has therefore no effect on the job creation decision. Of course, if the firm was risk-averse, the assumption that the firm bears all exchange rate risk would make a postponement of job creation in the presence of uncertainty even more likely.

VOL. 1 NO. 2

BELKE AND KAAS: EXCHANGE RATE VOLATILITY

25

Consider first the wage bargaining problem for a job created in period zero in which case the worker is hired for two periods. After the job is created (and the job creation cost is sunk), the (unconditional) expected net return of this job is equal to E0(S0) = 2p*–2w = 2π where π=p*−w denotes the expected return of a filled job per period (we abstract from discounting). Denoting the bargaining power of the worker by 0<β<1, the firm’s net return from the job created in period zero is9 E0(Π0) = (1–β)E0(S0) – c = 2(1–β)π – c .

(1)

In order to make the problem non-trivial, the expected return from job creation in period zero must be positive, that is we assume that 2(1–β)π–c > 0. Implicit in our model is the assumption that the firm and the worker sign a binding employment contract for two periods (zero and one). Hence they cannot sign a contract that allows for the possibility of job termination in the first period whenever the exchange rate turns out to be unfavorable. In period one (after realization of the exchange rate) the conditional expected surplus from job continuation is E1(S1)=π+e1 which may be negative if the exchange rate falls in period one below – π<0. In such circumstances, both the worker and the firm would benefit from termination. If a contract allowing for termination in period one could be signed, the unconditional expected surplus in period zero would be larger (consequently both the worker and the firm would prefer to sign such a contract).10 However, having in mind the interpretation of a rather short period length (a month, to be compatible with our empirical analysis), the assumption of a binding contract for two periods seems to be more appropriate. Of course, once a binding contract for two periods is signed, the worker always prefers continuation (since the contract wage exceeds the fallback wage), and the firm would incur losses if the exchange rate turns out to be unfavorable. We consider in Appendix 1 an alternative set-up which allows for the possibility of job destruction. It turns out in this case that uncertainty does not delay job creation, but rather that job destruction becomes more likely if uncertainty increases. Hence, the negative relationship between exchange rate variability and employment is robust to this variation. If the firm waits until period one it keeps the option open of whether or not to create a job. It will create a job only if the exchange rate realised during period one (and so expected for period two) is above a certain threshold level, or barrier, denoted by b. Given that an employment relationship in period one yields a return only during period two, this barrier to make the creation of the job just worthwhile is given by the condition that the (conditional) expected net return to the firm is zero: (1−β)(p* + b – w) − c = 0

or

b = c/(1−β) + w – p* = c/(1−β) – π . (2)

Whenever e1 ≥ b, the firm creates a job in period 1, and the conditional expected net return to the firm is E1(Π1) = (1–β)(π+e1)−c ≥ 0. Whenever e1 < b, the firm does not create a job in period one, and its return is zero. Hence, whenever both events occur

26

THE JOURNAL OF ECONOMIC ASYMMETRIES

DECEMBER 2004

with positive probabilities (that is whenever σ1 > b > −σ1)11, the unconditional expected return of waiting in period zero is given by: E0(Π1) = ((σ1 + b)/(2σ1))0 + ((σ1 – b)/( 2σ1))((1–β)(π + (σ1+b)/2) − c) ,

(3)

where the first term is the probability that it will not be worthwhile to create a job (in this case the return is zero). The second term represents the product of the probability that it will be worthwhile to create the job (because the exchange rate is above the barrier) and the average expected value of the net return to the firm under this outcome. Given condition (2) this can be rewritten as: E0(Π1) = (1–β) (σ1−b)2 / (4σ1).

(4)

This is an important result since it implies that an increase in uncertainty increases the value of the waiting strategy, because equation (4) is an increasing function of σ1.12 As σ1 increases it becomes more likely that it is worthwhile to wait until more information is available on the expected return during period two. At that point the firm can avoid the losses that arise if the exchange rate is unfavourable by not creating a job. This option not to create the job becomes more valuable with more uncertainty. The intuitive explanation is that waiting implies that the firm foregoes the expected return during period one, but it keeps the option not to create the job which is valuable if the exchange rate turns out to be unfavourable. The higher the variance the higher the potential losses the firm can avoid and the higher the potential for a very favourable realisation of the exchange rate, with consequently very high profits. It is now clear from (1) and (4) that a firm prefers to wait if and only if (1−β)(σ1–b)2 / (4σ1) > 2(1−β)π – c .

(5)

As the left hand side is increasing in σ1, the firm delays job creation if exchange rate uncertainty is large enough. The critical value at which (5) is satisfied with equality can be solved as 13 σ1* = 3π − c/(1−β) + 2 π(2π − c/(1 − β)) .

(6)

Whenever σ1>σ1*, firms decide to postpone job creation in period zero. Since σ1* is increasing in π (and thereby decreasing in the fallback wage w), decreasing in the cost of job creation c and decreasing in the worker’s bargaining power β, we conclude that a stronger position of workers in wage negotiations (reflected in a high fallback wage or in the bargaining power parameter) and higher costs of hiring raise the option value of waiting and make a postponement of job creation more likely.

VOL. 1 NO. 2

BELKE AND KAAS: EXCHANGE RATE VOLATILITY

27

Thus, the adverse impact of exchange rate uncertainty on job creation and employment should be stronger if the labor market is characterized by generous unemployment benefit systems, powerful trade unions, minimum wage restrictions or large hiring costs. That such features of the labor market are detrimental to employment is of course not surprising. The adverse impact of these features on employment has been confirmed empirically in various studies, and there are many other theoretical mechanisms explaining it (see for example Nickell, 1997, and Layard, Nickell and Jackman, 1991). What our simple model shows is that labor market rigidities also reinforce the negative employment effects of exchange rate uncertainty. Another important implication of the model is that only the current, shortterm uncertainty σ1 has an impact on the decision to wait. Future uncertainty, represented here by σ2, does not enter in the decision under risk neutrality. If one takes a fixed period, such as one month, the likelihood that job creation will be postponed to the end of that period depends only on the uncertainty during that period and not on future uncertainty. This implies that even short spikes in uncertainty as, for example, grasped by a contemporaneous uncertainty proxy in empirical investigations of the real option effect detected above, can have a strong impact on employment. Our crude model has abstracted from risk aversion. However, we believe that the basic conclusion is robust because a temporary spike in uncertainty would lead to higher discounting of expected future returns which makes a postponement of job creation even more likely under risk aversion. Moreover, the additional impact of risk aversion on job creation should be stronger under the realistic assumption that firms bear all the exchange rate risk In summary, we retain two conclusions from the model. First, even a temporary 'spike' in exchange rate variability can induce firms to wait with their creation of jobs (of course and for exactly this reason, the level of the exchange rate at the same time loses explanatory power). Second, the relationship between exchange rate variability and (un)employment should be particularly strong if the labour market is characterized by rigidities improving the bargaining position of workers. A stronger fallback position of workers raises the contract wage, lowers the net returns to firms and induces firms to delay job creation in the face of uncertainty. Our argument rests on the assumption that workers cannot be fired immediately if the exchange rate turns out to be unfavourable. Hence sunk wage payments are associated with the decision to hire a worker. These sunk costs and, consequently, the impact of uncertainty on job creation become more important if there are high firing costs. However, as we argue in Annex 4, even if there are no firing costs and if workers can be laid off at any point in time, exchange rate uncertainty should take a direct impact on job destruction. A more elaborate labor market model of job creation and job destruction (for example following the model of Pissarides, 2000, Chapter 3) might further clarify these issues, but we would expect that uncertainty has a negative effect on both job creation and destruction

28

THE JOURNAL OF ECONOMIC ASYMMETRIES

DECEMBER 2004

flows. In the empirical analysis, we therefore prefer to employ aggregate labor market indicators rather than more disaggregate job flow data.14 Our model is not detailed enough to have implications in terms of persistence. It has often been argued that (particularly in Continental Europe) there is hysteresis in unemployment; that is even temporary shocks can have permanent effects on unemployment. One channel through which hysteresis can arise is that the human capital of workers that have been fired depreciates rapidly so that they will not be able to find a new job at the old wage because they will have become less productive (see for example Blanchard and Diamond, 1994).15 If one interprets the set-up cost as relating to human capital this view could also be compatible with the model presented here. Hence even in our set-up there could be strong hysteresis. From an empirical perspective, exchange rate variability should thus enter any long-term relationship between employment and its determinants. However, some readers might have a strong perception that temporary shocks cannot have permanent effects. We do not want to take a stance on this issue here because it is not central to our analysis. The empirical results based on empirical tests of the significance of exchange rate variability in simple VARs which we present below are compatible with both views. 3. How to measure exchange rate variability? How should one measure exchange rate variability? We used a very simple measure: for each year of our sample 1973 to 2001 we calculated a standard deviation on the basis of twelve monthly observations of the first difference of the exchange rate. What kind of exchange rate did we take as the basis for our calculations? We used both the nominal effective rate of the US and the euro area (reconstituted for the past) and the bilateral euro-dollar rate. In order to have percentage changes we used the first difference of the raw numbers for the effective exchange rates as they are indices, with a base around 100. In the case of the bilateral euro/dollar rate we used the first difference of the natural logarithm. The historical series of the external exchange rate of Euroland was taken directly from the official sources, which calculate the average of bilateral exchange rates of the original 11 euro countries, with weights given by the non-euro trading partners. In order to convey an exact picture of our proceedings, the algorithm for the construction of the variability variables is described in Annex 3. We use monthly exchange rates to calculate volatility instead of daily (or other higher frequency) volatility because the required data were easier to obtain on a consistent basis for the entire sample period. Another reason to prefer this measure over more short-term alternatives (for example, daily variability) was that we are convinced that while the latter might be important for financial actors it is less relevant for decisions whether to export or to invest, which have a longer time horizon. The drawback of this decision was that we had to use annual data in order to have a meaningful measure of variability. We thus had only about 28 observations for each country but this turned out to be sufficient.

VOL. 1 NO. 2

BELKE AND KAAS: EXCHANGE RATE VOLATILITY

29

In principle, one could have used option prices to extract implicit forwardlooking volatilities, but options prices are generally available only for the US dollar and sometimes against the DM, and even then only for limited periods. Hence it would not have been possible to construct a measure of euro volatility on a consistent basis using option prices. We used actual exchange rate changes instead of only unanticipated ones, but at the monthly horizon the anticipated change is usually close to zero. Hence, actual and unanticipated changes should give the same results (Gros and Thygesen, 1992, 102, and Peeters, 1997, 5 ff.). An advantage of using monthly data is that price indices are available on a monthly basis so that one could use real exchange rates. We have preferred to use nominal rates in this first test since nominal and real exchange rates are usually highly correlated over a short-term horizon. The average variability (standard deviations) of the nominal effective exchange rate of the euro area was 1.13 (percent) for the post-1973 period, that of the US was much higher at 1.96 percent. Finally, the average volatility of the nominal dollar-euro exchange rate amounts to 2.35 percent. 4. Empirical analysis In this section we present and comment on the results of first tests of the importance of two measures of exchange rate variability (effective and bilateral) on two measures of labor market performance (changes in unemployment and employment growth) on both sides of the Atlantic. To start with a summary: exchange rate variability enters all equations with the expected sign and is in nearly all cases statistically significant. 4.1. Methodology Before commenting on the individual results we need to explain our methodology. In cases of doubt we always preferred taking differences since the disadvantages of differencing when it is not needed appear to us much less severe than those of failing to difference when it is appropriate. In the first case the worst outcome would be that the disturbances are moving average, but the estimators would still be consistent, whereas in the second case the usual properties of the OLS test statistics would be invalidated. All macroeconomic series were taken from the Ameco data set of the EC Commission. All exchange rate data were taken from the IMF (see in detail Annex 2). As a first step we present the results of some simple tests. We explain the first difference of the unemployment rate and the first difference of the index of employment by their own past and lags of our measure of exchange rate variability. The results, which are summarized below in the first row of Tables 1 and 2, are thus standard causality tests on the annual data used throughout this paper. The tables 1 and 2 summarize the regression results from bivariate VARs on annual data (1973-2001, sometimes shorter periods had to be used subject to data

30

THE JOURNAL OF ECONOMIC ASYMMETRIES

DECEMBER 2004

availability).16 The hypothesis tested is, as usual, that exchange rate variability does not have an influence on the two variables investigated here.17 All the results presented here are based implicitly on a comparison of two regression equations (notations chosen for consistency reasons, for a similar procedure see Belke and Gros 2001, 238 ff., and with an application to export shocks, see Belke and Gros 1999): N

DUEt = α0 +

∑α i =1

N

DUEt = α0 +

∑α i =1

i

⋅ DUE t −i + ut, and N

i ⋅ DUE t −i + ∑ β i ⋅ EXVt −i + ut, i =0

(8)

(9)

where DUEt stands for the change in the unemployment rate (between period t and t1), EXVt-i specifies the level of intra-European exchange rate variability (between period t-i and period t-i-1), ut represents the usual i.i.d. error term and N is the maximum number of considered lags (here according to Belke and Gros, 2001: 2 lags). Exchange rate variability (measured by an indicator as explained above in section 3) can then be said to “cause” unemployment if at least one ß, that is one of the coefficients on the past and contemporaneous (change in) exchange rate variability, is significantly different from zero. In other words, these tests measure the impact of the stationary level of exchange rate variability on changes in national unemployment rates once the autonomous movements in unemployment have been taken into account by including lagged unemployment rates among the explanatory variables. Thus, a significant effect (of whatever sign) implies that one can reject the hypothesis that exchange rate variability does not influence unemployment at the usual confidence levels. In order to be allowed to use the standard t-distribution for the purpose of model selection, one has to use changes of the unemployment rate as the level of this variable (as opposed to our measure of exchange rate variability) is clearly non-stationary. Substituting the change in employment (DEMPMAN) in the above setting describes our proceedings in the case of employment and investment instead of unemployment.

VOL. 1 NO. 2

BELKE AND KAAS: EXCHANGE RATE VOLATILITY

31

Table 1. Regression results based on the variability of the nominal effective exchange rate Euroland

Basic, best specification

US

Unemployment

Employment

Unemployment

Employment

0.61**

-0.63*

0.50**

-0.28

0.82***(-2)

-1.21*** (2)

0.57**

-0.56*

0.46**

-0.34

0.81*** (-2)

-0.51*

Robustness: additional variables

First differential of exchange rate

-1.22***

Spread (long-, short-term)

1.06***

-1.55***

0.31**

-0.55**

Real shortterm interest rate

1.01***

-1.52***

0.33*

-0.29

Change in real shortterm interest rate

1.00***

-1.50***

0.36**

-0.46*

Note:

Point estimates for the impact of exchange rate volatility are displayed together with their significance levels (***: 1 %; **: 5 %; *: 10 %). Numbers in brackets refer to the lags of the implemented volatility variable.

Table 1 shows our results using the level of effective nominal exchange rate variability and Table 2 the ones for the variability in the bilateral euro/dollar rate. For each of the two variables mentioned we first used as explanatory variables only their own past and lags of exchange rate variability. The results reported in the

32

THE JOURNAL OF ECONOMIC ASYMMETRIES

DECEMBER 2004

first row imply that exchange rate variability, whether measured by the standard deviation of the nominal effective rate or by that of the bilateral euro/dollar rate, has a significant impact. Table 2. Regression results based on the variability of the nominal bilateral euro/dollar exchange rate Euroland

US

Unemployment

Employment

Unemployment

Employment

0.41**

-0.57* (-1)

0.41**

-0.69** (-1)

First differential of exchange rate

0.40**

-0.54* (-1)

0.45**

-0.59** (-1)

Spread (long - short term)

0.45***

-1.01** (-1)

0.34**

-0.94*** (1)

Real short term interest rate

0.38**

-1.18* (-1)

0.37**

-0.69** (-1)

Change in real short term interest rate

0.55***

-1.48** (-1)

0.47***

-0.83** (-1)

Basic, best specification Robustness: additional variables:

Note:

-0.73** (-2)

Point estimates for the impact of exchange rate volatility are displayed together with their significance levels (***: 1 %; **: 5 %; *: 10 %). Numbers in brackets refer to the lags of the implemented volatility variable.

As exchange rate variability could be either caused by, or stand for some other macroeconomic variables, we also performed a series of robustness tests by adding • the level of the exchange rate,

VOL. 1 NO. 2

BELKE AND KAAS: EXCHANGE RATE VOLATILITY

33

• the spread between long- and short-term interest rates, and, • the (first difference of) real short-term interest rates. Only the coefficient estimate, its significance level and the lag order of exchange rate variability are displayed in the summary tables. The numbers in parentheses correspond to the lag order of exchange rate variability. If the impact effect is, for example, estimated to be lagged two years, this might indicate inflexibilities in the respective national labor market. The expected sign of the (change in) exchange rate variability is positive for (the changes in) the unemployment rate and negative for (the changes in) employment. The specification of the underlying equations is based on the usual diagnostics combined with the Schwarz Bayesian Information Criterion (SCH). The latter is chosen as our primary model selection criterion since it asymptotically leads to the correct model choice (if the true model is among the models under investigation, Lütkepohl, 1991). The regression which reveals the lowest SCH-value and at the same time fulfills the usual diagnostic residual criteria is chosen.18 As already stated above, the sample has been chosen to be 1973 to 2001 in order to exclude the Bretton Woods period of fixed exchange rates, which would have introduced structural breaks in the relationships. The procedure is exactly the same for each country. We never intervene to exercise a discretionary judgment. As usual, we add country specific dummies from time to time in order to account for possible breaks in the VAR relations. These dummies are added only if they improve the SCH statistics (higher informational contents even if a penalty for the extra dummy is taken into account) and do not lead to a rejection of the normality assumption of the residuals (Jarque, Bera, 1987). At the same time they should contribute to fulfill the criteria on the residuals, especially those on normality. However, none of our results is due to the implementation of these dummies. Most of the dummies were also economically meaningful (relating to the two oil crises, or the onset of EMU for Euroland) and most disappeared when policy variables were introduced in the robustness tests below. 4.2. Summary of results The results have to be read off the tables 1 and 2 as follows. In these tables, point estimates for the impact of exchange rate volatility are displayed together with their significance levels. For Euroland, the point estimate obtained from the first specification implies that a decrease of one percentage point in the variability (standard deviation) of the nominal effective exchange rate of the euro is associated during the same year with a decrease in the Euroland unemployment rate of nearly two thirds of a percentage point; and this is followed two years later by another reduction in the unemployment rate of 0.82 percentage points. We will comment only briefly on the impact coefficients because the longer-run effects depend of course on the dynamic behavior of the variables (Belke and Gros, 2001). The first upper right hand entry in Table 1 comes from a standard causality type regression whose results are reproduced in detail below in Table 3a in order to

34

THE JOURNAL OF ECONOMIC ASYMMETRIES

DECEMBER 2004

give a concrete example. This entry refers to the impact of the variability of nominal effective exchange rates on Euroland labor markets. The dependent variable in this case is represented by the change in the unemployment rate (DUREU). The depicted specification of the regression equation leads to the ‘best’ result in terms of the (lowest realization of) Schwarz criterion.19 Table 3a. Example regression for Euroland: unemployment rate on the variability of nominal effective exchange rates Dependent Variable: DUREU Method: Least Squares Sample: 1973 2001 Included observations: 29 Variable Coefficient Std. Error t-Statistic Prob. C -1.547368 0.450076 -3.438017 0.0022 DUREU(-1) 0.738169 0.129756 5.688894 0.0000 EXVNEEREU 0.614833 0.255707 2.404448 0.0246 EXVNEEREU(-2) 0.821263 0.269925 3.042559 0.0058 D83 -1.362151 0.520793 -2.615531 0.0155 D92 1.194570 0.453450 2.634402 0.0148 R-squared 0.677388 Mean dependent var 0.189655 Adjusted R-squared 0.607255 S.D. dependent var 0.703230 S.E. of regression 0.440709 Akaike info criterion 1.381129 Sum squared resid 4.467169 Schwarz criterion 1.664018 Log likelihood -14.02637 F-statistic 9.658632 Durbin-Watson stat 2.018440 Prob(F-statistic) 0.000045 A similar story emerges when one does the same test on the rate of employment growth defined as the first difference in the index of employment, that is roughly speaking the percentage change in the number of employed persons. Exchange rate variability had a significant impact on the European labor market from this angle as well. The regression result for the impact of the variability of nominal effective exchange rates for Euroland on the dependent variable employment (DEMPEU) is displayed in Table 3b (again we chose the ‘best’ fit in terms of lowest realization of the Schwarz criterion). Reverse causation does not appear plausible as mirrored by additional pairwise Granger-causality tests applied to exchange rate variability and the labor market variables used in the regressions. In addition, we rate the possibility that exchange rate variability at our high frequency was caused by slow-moving variables such as labor market rigidities or unemployment not being very high. Finally, there is evidence that exchange rates mainly react to financial rather than to real fundamentals like labor market variables (Canzoneri et al., 1996). If exchange rate volatility is largely noise (Rose, 1996, Flood and Rose, 1995), it does not make sense to treat this variable as endogenous and to regress it on labor market variables. Let us now turn to some robustness tests of the empirical results gained thus far.

VOL. 1 NO. 2

BELKE AND KAAS: EXCHANGE RATE VOLATILITY

35

Table 3b. Example regression for Euroland: employment growth on the variability of nominal effective exchange rates Dependent Variable: DEMPEU Method: Least Squares Sample: 1973 2001 Included observations: 29 Variable Coefficient Std. Error t-Statistic Prob. C 2.912109 0.733383 3.970791 0.0006 DEMPEU(-1) 0.617423 0.120250 5.134506 0.0000 EXVNEEREU -0.635770 0.343156 -1.852712 0.0774 EXVNEEREU(-1) -0.548227 0.337184 -1.625899 0.1182 EXVNEEREU(-2) -1.212938 0.325619 -3.725026 0.0012 D91 7.938697 0.637798 12.44705 0.0000 D92 -6.862967 1.183268 -5.800010 0.0000 R-squared 0.924131 Mean dependent var 0.757825 Adjusted R-squared 0.903439 S.D. dependent var 1.849384 S.E. of regression 0.574681 Akaike info criterion 1.936503 Sum squared resid 7.265691 Schwarz criterion 2.266540 Log likelihood -21.07930 F-statistic 44.66221 Durbin-Watson stat 1.727141 Prob(F-statistic) 0.000000 4.3. Robustness I: the impact of potential shock-absorbers The purpose of the following is to report the results of some tests for the robustness of the relationships found so far. We try to take into account the three most plausible ways in which exchange rate variability could stand for some other variable. For each hypothesis we then implement the same procedure based on the SCH criterion explained above. The three hypotheses we consider are: (i) Exchange rate variability is just a sign of a misalignment (that is a wrong level of the exchange rate). (ii) Exchange rate variability just reflects the stress caused by a tight monetary policy, the tightness of monetary policy being measured by the spread, the difference between long- and short-term interest rates. (iii) Exchange rate variability just reflects the stress caused by a tight monetary policy, but tight monetary policy is defined as high real short-term interest rates. On (i), a first possible reason for the significant negative (positive) correlation of exchange rate variability with (un-) employment might be that this volatility just stands for misalignments of the real exchange rate. The basic argument is simple: the dollar (respectively the euro, or its main component, the DM) was very strong when it was also variable. This argument could also be turned on its head because one suspects that the dollar was variable when it was very weak. But it needs to be addressed because it represents a popular explanation for the results we obtain.

36

THE JOURNAL OF ECONOMIC ASYMMETRIES

DECEMBER 2004

In order to take this hypothesis into account, we added the first difference (the level is not stationary) of the (nominal) effective exchange rate (NEER) in the regressions displayed in Tables 1 and 2, second rows. Note again, that point estimates for the impact of exchange rate volatility are displayed together with their significance levels. The results suggest that this hypothesis does not hold a lot of explanatory power as the addition of the level of the exchange rate does in no case change the magnitude or significance level of the coefficient of exchange rate variability. Except for the case of the US (employment), the latter remains highly significant. On (ii), transatlantic exchange rate variability could also just be the result of tight monetary policy pursued on either side. The hypothesis is that a restrictive monetary policy leads to employment losses in the short term, and that this is exclusively assigned to exchange rate variability in Tables 1 and 2. However, this problem of identification can be reduced by explicitly adding a variable that indicates the tightness of the national monetary policy to the equation. We use the spread (long- minus short-term interest rates) as a first indicator. This control variable actually improves the performance of the equation overall and has the additional advantage of eliminating the two lagged effects that appear for Euroland in some cases. On (iii), adding only the real short term interest rate to the equation also does not change the results in the sense that the coefficient on exchange rate variability remains significant. We used both the level and the first difference of this control variable because it was not clear whether it is stationary or not. However, as the last two rows of tables 1 and 2 show, the results are virtually identical whether one uses the level or the first difference (at least if one uses nominal effective exchange rates). For Euroland we thus find that in all equations exchange rate variability is significant and enters with the expected sign. For the US there are, however, more entries in the unemployment column. It is interesting to note that, by contrast, for Euroland the impact on employment seems to be stronger. Taking the strong evidence in favor of eurosclerosis in some larger countries of the euro-zone, such as Germany, into account, this strong result is totally in line with the labor market model developed in Section 2.20 4.4. Robustness II: the hedging issue Some sceptics might raise the methodological caveat that hedging is ignored in our considerations and hence, observed exchange rate variability does not correspond to uncertainty. However, hedging is not ignored in our model but either implicitly taken into account in the model or neglected due to theoretical considerations. Let us start our argument with the observation that hedging operations clearly have an impact on costs and thus have an impact on employment similar to a tariff (dell’Ariccia 1998, 8, and Edison and Melvin 1990, 20). Any efforts to include the possibility of hedging via futures contracts the EXR risk in our theoretical model would in fact boil down to an inclusion of the hedging costs in the fixed (sunk) costs

VOL. 1 NO. 2

BELKE AND KAAS: EXCHANGE RATE VOLATILITY

37

and, at the same time, to a substitution of the realizations of the expected returns by certain returns which, however, take the same values as the expected ones. Hence, the pattern of the main results of the model should stay the same. In this sense, hedging possibility is already included in the model. However, in the following we also deliver some important arguments speaking in favor of our hypothesis that the EXR risk is not or cannot be hedged (completely). First, we already indicated that some sceptics might be inclined to call our measure of EXR uncertainty unconditional. However, the fact that we cannot reject empirically an impact of even short-term spikes of our (allegedly unconditional) measure of EXR volatility on (un-) employment, reveals that the hedging facilities might in fact be inadequate. “This may exhibit inadequacy of hedging facilities” (Morsink and Molle 1991, 47, with respect to the impact of EXR volatility on FDI). Second, physical re-allocation, that is outsourcing, is often claimed to lead to diminished employment at home. In this sense, physical hedging could be interpreted as export of jobs. Hence, in this view, hedging itself would be responsible for a negative (positive) impact of EXR volatility on (un-) employment (Kulatilaka and Kogut 1996, Viaene and de Vries 1992). Third, the hypothesis that the exchange rate risk can be insured completely since there is no risk premium in equilibrium is not corroborated throughout empirical studies (see, for example, Malliaropoulos 1995). “Forward markets, where traders hedge against exchange rate uncertainty, have repeatedly failed to predict correctly even the direction of exchange rate changes” (Kumar, Whitt 1992, 27). As a stylized fact, EXR risk is not fully covered by ordinary futures contracts. One reason is that futures markets are usually thinner than spot markets. Hence, in periods of stress on FX markets, the corresponding bid-ask spreads become larger than usual (for example, Stokman 1995, 42 f.). Large firms in contrast to smaller ones prefer futures contracts which do not have an impact on the investment calculus of employers (see our argument above). Finally, one has to take into account that future markets are incomplete for maturities over 1 year (note that our model assumes two periods). Although large enterprises have the possibility to incur debt in foreign currency, this technique leads to even higher costs and is suitable only for large firms with access to international capital markets or to markets for foreign currency swaps or barter deals (for example, Edison and Melvin 1990, 20, and Stokman 1995, 43). However, employment decisions in both considered economies are not solely made by large firms or even multi-nationals. In the light of the above arguments, we would thus still like to argue that observed exchange rate variability closely corresponds to uncertainty. 5. Conclusions Our main policy conclusion is that reducing exchange rate variability in the two dominant G3 economies should bring substantial benefits. The data from the past suggest that exchange rate variability had a statistically significant negative impact

38

THE JOURNAL OF ECONOMIC ASYMMETRIES

DECEMBER 2004

on unemployment and employment in Euroland. For the US, the evidence points only to an impact on unemployment. This weaker result is on the one hand probably due to more flexibility in the labor markets and, hence, to a minor importance of hiring and firing costs for hiring and firing decisions in the US. However, volatility in the bilateral rate and in the nominal effective rates seems to matter. On the other hand, the potential exposure of the US to exchange rate variability as compared to Euroland is significantly weaker. If one looks at the share of trade in national income (Gros and Weiner 2000) it becomes obvious that while Euroland is in the aggregate less open than its constituent members, it is substantially more open than the United States.21 A final reason why the US is less exposed by EXR fluctuations than the euro area might be the fact that US corporations hedge much more than the European ones. The great importance of the 1 trillion foreign exchange daily market cannot be ignored in this respect.22 Therefore, we would argue that volatility matters because employment decisions (as investment decisions) have some degree of irreversibility. Job creation is discouraged by higher exchange rate variability, and the effect should be more pronounced when labor markets are “rigid”. The estimated effect of volatility on (un)employment might appear to be economically small. In fact, a decrease of one percentage point in the standard deviation of Euroland’s nominal effective exchange rate (which amounts to abandoning all volatility) reduces unemployment by only half a percentage point. However, we should point out that this is only the impact effect in the first year, whereas there are also substantial lag effects in some regressions. Moreover, hysteresis in unemployment could trigger economically significant and persistent long-run effects. The standard dichotomy between structural unemployment and nominal volatility does not hold (see Ramey and Ramey 1995 in the context of growth and business cycle volatility). However, it has to be noted that an alternative labor market adjustment to exchange rate volatility is to diversify the production capacity via FDI (Aizenman 1994), resulting in potentially similar employment impacts. A common argument against reducing exchange rate variability is the position that volatility must have a valve somewhere else. In other words, could the gains from suppressing exchange rate variability that are suggested by our results be lost if the volatility reappear elsewhere, for example in higher interest rate variability? We would argue that recent research is suggestive in this respect. Seen on the whole, the existing literature is skeptical about the “squeeze the balloon” theory, that is a trade off between exchange rate volatility and the volatility of other variables. Rose (1996), for example, shows that official action can reduce exchange rate variability even holding constant the variability of fundamentals such as interest rates and money. Co-ordination between the Fed and the ECB could thus keep the dollar-euro volatility under control. This view is supported by results of Flood and Rose (1995) who show that there is no clear tradeoff between exchange rate volatility and macroeconomic stability. Furthermore, Jeanne and Rose (1999) develop a model of a foreign exchange market with an endogenous number of noise

VOL. 1 NO. 2

BELKE AND KAAS: EXCHANGE RATE VOLATILITY

39

traders and multiple equilibria of high and low exchange rate volatility. In their model monetary policy can be used to lower exchange rate volatility without affecting macroeconomic fundamentals. In the same vein, Canzoneri et al., (1996) show that exchange rates do not generally move in the direction one would expect if they were to offset shocks. Finally, the gains from suppressing intra-ERM exchange rate variability by EMU have not been substituted by a higher effective variability of the euro. Much of the commentary on the euro equates the depreciation of the euro with exchange rate variability. However, it is not widely appreciated that, in reality, the fall of the euro has occurred in a relatively smooth manner. Short-term variability, as measured by the volatility of monthly percent changes has actually been rather low compared to the past, as shown by Gros et al., (2000). The variability of the precursor of the euro, the ecu, has on average been about 1.1 percent (per month) since 1980. In 1999, this fell to 0.7 percent. A visual inspection of the plots of the various volatility measures in Annex 1 corroborates this conclusion. We realize that our results are preliminary, not least because the questions posed in this paper have not been posed in this way in the literature so far. The limited number of observations we have, given the annual data we use, are a further reason to be cautious.23 Many readers might sympathize with the point of view that exchange rate variability is usually not connected with variability in the fundamentals and thus undesirable. Unfortunately, concrete action to reduce exchange rate variability at least among the G-3 is either impossible or politically unacceptable (Mundell 2000, 2000a). The same is valid with respect to labor market deregulation in the Euro zone. However, both measures would according to our model be complementary to eliminate impacts of euro/dollar volatility on labor markets. The main focus of the paper is to give models of the exchange rate volatility/labor market channel a stronger theoretical background and to illustrate the main findings by first simple regressions. However, much further work is needed to corroborate our first preliminary empirical results so that they can be used as a basis for concrete policy recommendations. In particular, one should concentrate on the implications for the debate on the design of EU-US monetary relations and especially on the role one believes the exchange rate should play in monetary policy, that is the desirability of influencing the exchange rate.

40

THE JOURNAL OF ECONOMIC ASYMMETRIES

DECEMBER 2004

Annex 1: Exchange Rate Variability - Different Operationalizations 2.4

5

2.0

4

1.6

3

1.2

2

0.8

1

0.4

0 1975

1980

1985

1990

1995

2000

1975

1980

EXVNEEREU

1985

1990

1995

2000

1995

2000

EXVNEERUS

4.0

2.0

3.5

1.8 1.6

3.0

1.4 2.5 1.2 2.0

1.0

1.5

0.8

1.0

0.6 1975

1980

1985

1990

1995

2000

EXVNERDOLLECU

5 4 3 2 1 0 1980

1985

1990

EXVREERUSULC

1980

1985

1990

EXVREEREUULC

6

1975

1975

1995

2000

VOL. 1 NO. 2

BELKE AND KAAS: EXCHANGE RATE VOLATILITY

41

Annex 2: Data Annual data: a) Euroland data UREU: EU-11 unemployment rate by Eurostat, Eurostat Statistics CD ed. 2001. Labor markets euro-zone (EU-11: BE,DE,ES,FR,IE,IT,LU,NL,AT,PT,FI) EMPEU: Total employment EU-11 in thousands (AMECO file, EU Commission) DEMPEU = LOG(EMPEU)-LOG(EMPEU(-1)) INTLEU: Nominal long-term interest rate EU-11 (AMECO) INTSREU: Real short-term interest rate (deflator GDP) EU-11 (AMECO) DINTSREU = INTSREU - INTSREU(-1) INTSEU: Nominal short-term interest rate EU-11 (AMECO) SPREADEU: EU-11 yield curve (AMECO) b) United States data URUS: US-Unemployment rate EUROSTAT definition (AMECO) EMPUS: Total employment EU-11 in thousands (AMECO DEMPUS = LOG(EMPUS)-LOG(EMPUS(-1)) INTLUS: Nominal long-term interest rate US (AMECO) INTSRUS: Real short-term interest rate (deflator GDP) US (AMECO) DINTSRUS = INTSRUS - INTSRUS(-1) (no logs because of non-positive numbers) INTSUS: Nominal short-term interest rate US (AMECO) SPREADUS: US-yield curve (AMECO) c) Exchange rate variability data (Euroland and United States) EXVNEEREU: Exchange rate variability from NEEREU EXVREEREUULC: Exchange rate variability from REEREUULC EXVNERDOLLECU: Exchange rate variability from NERDOLLECU EXVNEERUS: Exchange rate variability from NEERUS EXVREERUSULC: Exchange rate variability from REERUSULC

42

THE JOURNAL OF ECONOMIC ASYMMETRIES

DECEMBER 2004

Monthly data: Exchange rates (source: International Monetary Fund, IFS) NEEREU

EURO AREA IFS 163..NEUZF... NEER FROM ULC

REEREUULC

EURO AREA IFS 163..REUZF... REER BASED ON RNULC

NERDOLLECU USA

IFS 111..EB.ZF... US $/ECU RATE: PERIOD AV.

NEERUS

USA

IFS 111..NEUZF... NEER FROM ULC

REERUSULC

USA

IFS 111..REUZF... REER BASED ON RNULC

Annex 3: Algorithm to calculate the exchange rate variability series SMPL 1960.1 2001.12 FOR %EX NEEREU NEERUS REEREUCPI REEREUULC REERUSCPI REERUSULC GENR EXV%EX = NA FOR !1=0 to 492 STEP 12 SMPL 1960.1+!1 1960.12+!1 GENR EXV%EX=SQR(@VAR(D(%EX))) NEXT NEXT SMPL 1960.1 2001.12 FOR %EX NERDOLLECU GENR EXVNERDOLLECU = NA FOR !1=0 to 492 STEP 12 SMPL 1960.1+!1 1960.12+!1 GENR EXVNERDOLLECU=SQR(@VAR(D(LOG(NERDOLLECU))*100)) NEXT NEXT SMPL 1960.1 2001.12

VOL. 1 NO. 2

BELKE AND KAAS: EXCHANGE RATE VOLATILITY

43

Annex 4: Supplement to Section 2 In this annex, we consider the scenario of a labor market in which the firm and the worker can sign a contract only for one period and keep the option to terminate the work relationship whenever it becomes unprofitable. In period 1, the conditionally expected surplus of job continuation is π+e1 which is positive whenever e1>−π. Hence, whenever uncertainty is large enough (σ1>π), there is job destruction in period 1 with probability (σ1−π)/(2σ1). The (unconditional) expected net return to the firm from a job created in period zero (and with the option of destruction in period one) is therefore E0(Π0) = ((1−β)π − c) + ((σ1 – π)/2σ1)0 + ((σ1 + π)/2σ1)(1−β)(π + (σ1 – π)/2)) , (10) where the first term is the expected return from the job in period one, whereas the second and third term represent the expected surplus from the job in period two (after destruction or after continuation in period one) under the assumption σ1>π. If σ1<π, the job would never be destroyed, and the expected net return is, as before, 2(1 − β)π − c . Hence, after rearranging (10), the expected net return from a job created in period zero can be written

 2(1 − β)π − c , if σ1 < π , 2 (1 − β) ( π + (σ1 + π) /(4σ1 ) ) − c , if σ1 ≥ π .

E0(Π0) = 

On the other hand, if the firm waits until period 1, the (unconditional) expected net return is, as in Section 2,

 max(0,(1 − β)π − c) , if σ1 < | π − c/(1 − β)| ,

E0(Π1) = 

2 (1 − β)(σ1 + π − c/(1 − β)) /(4σ1 ) , if σ1 ≥ | π − c/(1 − β)|

It is now easy to see that the firm never delays job creation. First, if σ1 ≤ | π − c/(1 − β)| <π, the firm never destroys a job in period 1, and so we have E0(Π0)>E0(Π1). Second, if σ 1 ≥ π , the condition E0(Π0)>E0(Π1) means that 4σ1(π−c/(1−β)) + (σ1+π)2 > (σ1+π−c/(1−β))2 which turns out to be equivalent to (2(1−β)π−c)(c/(1−β)+2σ1)>0 and which is satisfied because of our assumption 2(1−β)π−c>0. Hence, the firm does not delay job

44

THE JOURNAL OF ECONOMIC ASYMMETRIES

DECEMBER 2004

creation also in this case. Finally, if π−c/(1−β)< σ1< π, the condition E0(Π0)>E0(Π1) means that 4σ1(2(1−β)π−c) − (1−β)(σ1+π−c/(1−β))2 > 0 . But since this inequality is satisfied at the boundaries σ1=π and σ1=π−c/(1−β)and since the left hand side is a concave function of σ1, the inequality is also satisfied in the interval π−c/(1−β)< σ1< π. Hence, firms always prefer to create a job in period zero, and so exchange rate uncertainty has no impact on job creation. However, since there is job destruction with probability (σ1−π)/(2σ1) (whenever σ1>π), the probability of job destruction is increasing in uncertainty. Hence, there is also a negative impact of exchange rate uncertainty on employment in this case. Moreover, this effect is more pronounced if the worker’s fallback wage is higher (if π is smaller). Therefore, the basic conclusions of Section 2 remain valid. Notes Angsar Belke, University of Hohenheim, D-70593 Stuttgart, Germany, E-mail: [email protected] Leo Kaas, University of Vienna, Hohenstaufengasse 9, 1010 Vienna, Austria. E-mail: [email protected] Leo Kaas would like to thank the Austrian Science Fund (FWF) for financial support. Ansgar Belke gratefully acknowledges financial support from the Viennese Chamber of Commerce Science Fund. We are grateful to Joshua Aizenman, Daniel Gros and Erich W. Streissler as well as two anonymous referees for valuable comments. 2 See, for example, IMF (1984) and Côté (1994). However, Rose (1999) reports a small significant negative effects of exchange rate volatility on trade based on panel data for 186 countries. 3 Dornbusch (1987). See also Krugman (1989); CEC (1995) documents the more recent European data. 4 For an analysis of the costs of intra-European variability for European labor markets which was suppressed by EMU see Belke and Gros (2001). Their results have only recently been corroborated to a large extent by Mueller and Buscher (1999) and Buscher and Stirboeck (2000). 5 These are the two politically most important variables of the indicators linked to popular explanations of the impact of financial volatility on the real sector (Dixit, 1989, Aizenman and Marion, 1996, Ramey and Ramey, 1995). On the other hand, the lack of comparable data for European countries prevents us from testing the effects on job creation and job destruction flows directly. 6 Aizenman and Marion (1999) provide further empirical evidence on a negative relation between various volatility measures and private investment. They argue that increasing volatility has a negative impact on investment if investors are disappointment-averse. Moreover, in the presence of credit constraints, realized 1

VOL. 1 NO. 2

BELKE AND KAAS: EXCHANGE RATE VOLATILITY

45

investment is lower on average when investment demand is more volatile, since credit constraints are more often binding. Real impacts of volatility are also confirmed by Ramey and Ramey (1995). 7 Although not undisputed, there is some support for the PTM hypothesis, at least for developed countries. We deliberately employ this hypothesis for the purpose of our illustrative model. 8 An interesting aspect of this crude model is that it does not contain an often used assumption, namely that the uncertainty is resolved at the end of the first period. In reality uncertainty is usually not resolved, but persists. In a model with an infinite horizon this could imply that the same decision represents itself every period in the same way. For example, the European Monetary Union constitutes an exception to the rule that uncertainty just continues in the sense that the start of EMU should definitely eliminate uncertainties about the economic environment. 9 Formally, the wage bargain leads to a wage rate maximizing the Nash product (2w-2w)β(2p*-2w)1-β whose solution is w=(1-β)w+βp*, and hence the expected net return for the firm is 2p*-2w-c=(1-β)(2p*-2w)-c. 10 Of course, such a flexible contract implies that some exchange rate risk is shared between the worker and the firm. However, the reason why they both benefit is not the risk-sharing aspect, but the fact that the flexible contract excludes continuation of unprofitable work relationships. 11 We do not a priori restrict the sign of the barrier b. Hence one of these conditions is automatically satisfied, whereas the other is satisfied only if uncertainty is large enough. 12 Formally, this results from the fact that equation (4) is only valid whenever σ1 exceeds b (otherwise the exchange rate could never exceed the barrier and the firm never creates a job in period 1) and whenever −σ1 is lower than b (otherwise the exchange rate could never fall below the barrier and the firm always creates a job in period one). 13 The other (smaller) solution to this equation is less than |b| and is therefore not feasible. 14 Klein, Schuh and Triest (2000) investigate the impact of exchange rate movements on job flows in the US. They find a response of job destruction to dollar appreciation, whereas job creation does not respond significantly to depreciations. This result reflects the asymmetric responses of job creation and destruction to aggregate shocks that have been detected in other studies. It does not contradict our conclusions, however, since job creation might just respond to exchange rate volatility rather than to actual appreciations or depreciations. 15 Other mechanisms of hysteresis are the insider-outsider hypothesis of Blanchard and Summers (1986), the hypothesis of declining search-intensities of long-term unemployed (Layard, Nickell and Jackman, 1991) or the idea that credit market

46

THE JOURNAL OF ECONOMIC ASYMMETRIES

DECEMBER 2004

imperfections may lead to slower adjustment to technological innovations and delay job creation (Acemoglu, 2000). 16 The individual regression results are available on request. 17 We thus use VARs in first differences of the respective real variables. Since we classify all real variables as integrated of order one we feel justified to deviate from the usual specification of an Augmented Dickey-Fuller test (including a drift term) only by neglecting the (insignificant) lagged endogenous level variable. The significance of the coefficient estimates of the lags of the changes in the real variables and of the indicator of exchange rate variability can then be judged on the basis of the usual standard normal respectively the asymptotic values of the studentt-distribution. Cf. Belke and Gros (2001) and Haldrup (1990), 31 f. 18 However, one important precondition for their application is the same number of observations for the alternative specifications. See Banerjee et al. (1993, 286), Mills (1990, 139), and Schwarz (1978). 19 Samples being the same throughout. 20 We enacted our regression analysis for measures of real exchange rate variability (see annex) as well. These measures are highly correlated with our measures of nominal exchange rate variability and led to nearly the same results (which are available by request). 21 Hence, Robert Mundell (2000) argues that it would be a great mistake to believe that the closed nature of the three big blocs of the G-3 would make exchange rates less important, or that the dollar-euro rate can be treated with ‘benign neglect’. In addition, total bilateral trade between Euroland and the United States is the most important bilateral trade relationship in the world indicating the relative importance of the US dollar/euro exchange rate. He argues that given the large degree of inflation convergence achieved the long term thrust of monetary policy is actually very similar throughout the G-3, it should be possible to agree on a common line that makes it possible to contemplate joint action to reduce excessive exchange rate variability. 22 We are grateful to one anonymous referee pointing out this argument to us. 23 But we are encouraged by the extent to which previous results for intra-European exchange rate variability have been able to withstand the numerous robustness tests conducted by ourselves and by recent studies (Mueller and Buscher, 1999, Buscher and Stirboeck, 2000). References Acemoglu, D., (2000), “Credit Market Imperfections and Persistent Unemployment”, NBER Working Paper 7938, Cambridge, MA. Aizenman, J., (1994), “Monetary and Real Shocks, Productive Capacity and Exchange Rate Regimes”, Economica 61, 407-434.

VOL. 1 NO. 2

BELKE AND KAAS: EXCHANGE RATE VOLATILITY

47

Aizenman, J., and Marion, N.P., (1996), “Volatility and the Investment Response”, NBER Working Paper 5841, Cambridge, MA. Aizenman, J., and Marion, N.P., (1999), “Volatility and Investment: Interpreting Evidence from Developing Countries”, Economica 66, 157-179. Andersen, T. M. and Sorensen, J.R., (1988), “Exchange Rate Variability and Wage Formation in Open Economies”, Economics Letters 28, 263-268. Banerjee, A., Dolado, J., Galbraith, J.W., and Hendry, D., (1993), Co-Integration, Error Correction, and the Econometric Analysis of Non-Stationary Data, Oxford. Belke, A. and Goecke, M., (1998), “A Simple Model of Hysteresis in Employment under Exchange Rate Uncertainty”, Scottish Journal of Political Economy 46, 260-286. Belke, A. and Gros, D., (1999), “Estimating the Costs and Benefits of EMU: The Impact of External Shocks on Labor Markets”, Weltwirtschaftliches Archiv 135, 1-48. Belke, A. and Gros, D., (2001), “Real Impacts of Intra-European Exchange Rate Variability: A Case for EMU?”, Open Economies Review 12, 231-264. Blanchard, O.J. and Diamond, P.A., (1994), “Ranking, Unemployment Duration and Wages”, Review of Economic Studies, 57, 417-434. Blanchard, O.J. and Summers, L., (1986), “Hysteresis and the European Unemployment Problem”, NBER Macroeconomics Annual, 15-78. Buscher, H.S. and Stirboeck, C. (2000), “Exchange Rate Volatility Effects on Labour Markets”, Intereconomics – Review of International Trade and Development, 35, 9-22. Canzoneri, M.B., Javier, V. and Viñals, J., (1996), “Do Exchange Rates Move to Address National Imbalances?”, CEPR Discussion Paper 1498, London. Chinn, D.M. and Miller, P.J., (1998), „Fixed vs. Floating Rates: A Dynamic General Equilibrium Analysis”, European Economic Review 42, 1221-1249. Commission of the European Communities (1995), The Impact of Currency Fluctuations on the Internal Market, Brussels. Côté, A. (1994), “Exchange Rate Volatility and Trade – A Survey,” Bank of Canada Working Paper 94-5, Ottawa. Darby, J., Hughes Hallett, A., Ireland, J. and Piscitelli, L., (1998), “The Impact of Exchange Rate Uncertainty on the Level of Investment”, CEPR Discussion Paper 1896, London. De Grauwe, P., (1987), “International Trade and Economic Growth in the European Monetary System”, European Economic Review 3l, 389-398. De Grauwe, P. and Skudelny, F., (1997), “The Impact of EMU on Trade Flows”, Katholieke Universiteit Leuven Department of Economics Discussion Paper 133, Leuven.

48

THE JOURNAL OF ECONOMIC ASYMMETRIES

DECEMBER 2004

Dell’Ariccia, G., (1998), “Exchange Rate Fluctuations and Trade Flows - Evidence from the European Union”, IMF Working Paper 107, Washington D.C. Devereux, M.B. and Engel, C., (1998), “Fixed vs. Floating Exchange Rates: How Price Setting Affects the Optimal Choice of Exchange-Rate Regime”, NBER Working Paper 6867. Dixit, A., (1989), “Entry and Exit Decisions under Uncertainty”, Journal of Political Economy 97, 620-638. Dornbusch, R., (1987), “Exchange Rates and Prices”, American Economic Review 77, 93-107. Edison, H. J. and Melvin, M., (1990), “The Determinants and Implications of the Choice of an Exchange Rate System”. In: Haraf, W.S., Willett, T.D. (eds.), Monetary Policy for a Volatile Global Economy, 1-44, AEI Press, Washington D.C. Flood, R.P. and Rose, A.K., (1995), “Fixing Exchange Rates – A Virtual Quest for Fundamentals”, Journal of Monetary Economics 36, 3-37. Fountas, S. and Aristotelous, K., (1999), “Has the European Monetary System Led to More Exports? Evidence from Four European Union Countries”, Economics Letters 62, 357-363. Gros, D. and Thygesen, N., (1992), European Monetary Integration, Harlow, New York. Gros, D. and Weiner, J., (2000), “A Note on Measures of Openness or Exposure to External Shocks”, Manuscript Prepared for the 1st Workshop “Designing EUUS Atlantic Monetary Relations”, Centre for European Policy Studies, Brussels. Gros, D., Davanne, O., Emerson, M., Mayer, T., Tabellini, G. and Thygesen, N., (2000), “Quo Vadis Euro? – The Cost of Muddling Through,” Second Report of the CEPS Macroeconomic Policy Group, Brussels, 15 May. Haldrup, N., (1990), “Testing for Unit Roots with a Maintained Trend When the True Data Generating Process Is a Random Walk with Drift”, Institute of Economics Memo 1990-22, University of Aarhus. International Monetary Fund (1984), “Exchange Rate Volatility and World Trade”, IMF Occasional Paper 28, July, Washington. Jarque, C. M. and Bera, A.K., (1987), “A Test for Normality of Observations and Regression Residuals”, International Statistical Review 55, 163-172. Jeanne, O. and Rose, A.K., (1999), “Noise Trading and Exchange Rate Regimes”, NBER Working Paper no. 7104, Cambridge, MA. Klein, M.W., Schuh, S. and Triest, R.K., (2000), “Job Creation, Job Destruction, and the Real Exchange Rate”, NBER Working Paper no. 7466, Cambridge, MA. Krugman, P., (1989), Exchange Rate Instability, Cambridge/MA: MIT Press.

VOL. 1 NO. 2

BELKE AND KAAS: EXCHANGE RATE VOLATILITY

49

Kulatilaka, N. and Kogut. B., (1996), “Direct Investment, Hysteresis, and Real Exchange Rate Volatility”, Journal of the Japanese and International Economies 10, 12-36. Kumar, V. and Whitt, J.A., (1992), “Exchange Rate Variability and International Trade”, Federal Reserve Bank of Atlanta Economic Review 37, 17-32. Layard, R., Nickell, S. and Jackman, R., (1991), Unemployment, Macroeconomic Performance and the Labour Market, Oxford: Oxford University Press. Lütkepohl, H., (1991), Introduction to Multiple Time Series Analysis, New York: Springer. Malliaropoulos, D., (1995), “Conditional Volatility of Exchange Rates and the Risk Premia in the EMS”, Applied Economics 27, 117-123. Mills, T.C., (1990) Time Series Techniques for Economists, Cambridge et al.: Cambridge University Press. Morsink, R.L.A. and Molle, W.T.M., (1991), “Direct Investments and Monetary Integration”. In: Background Studies for European Economy No. 44, One Market, One Money, 36-55. Mueller, C. and Buscher, H.S., (1999), “Exchange Rate Volatility Effects on the German Labor Market: A Survey of Recent Results and Extensions”, IZA Discussion Paper No. 37, Institute for the Study of Labour (IZA), Bonn. Mundell, R., (2000), “Threat to Prosperity”, in: The Wall Street Journal, Thursday, March 30. Mundell, R., (2000a), “A Reconsideration of the Twentieth Century”, in: American Economic Review 80, 327-340. Neumeyer, P., (1998), “Currencies and the Allocation of Risk: The Welfare Effects of a Monetary Union”, American Economic Review 88, 246-259. Nickell, S., (1997), “Unemployment and Labor Market Rigidities: Europe versus North America”, in: Journal of Economic Perspectives 11, 55-74. Peeters, M., (1997), “Does Demand and Price Uncertainty Affect Belgian and Spanish Corporate Investment?”, Dutch National Bank-Staff Reports 13, Amsterdam. Pindyck, R.S., (1991), “Irreversibility, Uncertainty, and Investment”, Journal of Economic Literature 29, 1110-1148. Pissarides, C., (2000), Equilibrium Unemployment Theory, 2nd ed., Cambridge/MA. Ramey, G. and Ramey, V.A., (1995, “Cross-Country Evidence on the Link Between Volatility and Growth”, American Economic Review 85, 1138-1151. Rose, A.K., (1996), “After the Deluge: Do Fixed Exchange Rates Allow InterTemporal Volatility Trade-Offs?”, International Journal of Finance and Economics 1, 47-54. Rose, A.K., (1999), “One Money, One Market: Estimating the Effect of Common Currencies on Trade”, NBER Working Paper no. 7432, Cambridge, MA.

50

THE JOURNAL OF ECONOMIC ASYMMETRIES

DECEMBER 2004

Rowthorn, R., (1999), “Unemployment, Wage Bargaining and Capital-Labour Substitution”, Cambridge Journal of Economics 23, 413-425. Sapir, A. and Sekkat, K., (1990), “Exchange Rate Volatility and International Trade”. In: de Grauwe, P. and Papademos, L., (eds.), The European Monetary System in the 1990s, Longman, 182-198. Stokman, Ad C. J. (1995), “Effect of Exchange Rate Risk on Intra-EC Trade”, De Economist 143, 41-54. Viaene, J. and de Vries, C.g., (1992), “International Trade and Exchange Rate Volatility”, European Economic Review 36, 1311-1321. Schwarz, G. (1978), “Estimating the Dimension of a Model”, Annals of Statistics 6, 461-464. Sorensen, J.R., (1992), “Uncertainty and Monetary Policy in a Wage Bargaining Model”, Scandinavian Journal of Economics 94, 443-455.